Computer vision image processing techniques are increasingly more frequently employed to study and characterize complex size, shape and features of particles. Photographic image acquisition techniques make it possible to capture and analyse flocs’ images online (Chakraborti et al. 2003). The distinct effect of coagulant dosage on aggregates’ features was detected by Lin et al. (2008) in kaolin water suspension. Authors used wet scanning electron microscopy to obtain the morphology of flocs. Jin (2005) used high-resolution digital camera to study the influence of temperature on flocs properties under different coagulant dosages for river water coagulation. The relations between the projected area of particles and coagulant dosages were found. Wang et al. (2011) investigated the changes in flocs’ characteristics due to different dosages and coagulation pH in humic acid suspension, obtaining images by digital CCD camera.
Image analysis of particles is a fairly old (Tambo & Watanabe 1979b) and known technique to be used in different water treatment applications, for instance: activated sludge processing (Alves et al. 2000; Amaral & Ferreira 2005; da Motta et al. 2001; Dagot et al. 2001;
Jenné et al. 2006; Mesquita et al. 2011), aggregates settling velocity measurements (Vahedi &
Gorczyca 2012; 2014), membrane fouling (Mendret et al. 2007), natural organic matter removal (Xiao et al. 2011). Image analysis applied to coagulation and aggregation of particles was also widely studied in different waters: clay (kaolinite) suspension (Tambo & Watanabe 1979a; Xiao et al. 2011; He et al. 2012; Lin et al. 2008), mineral suspensions (Gorczyca &
Ganczarczyk 1996), latex particles suspension (Chakraborti et al. 2003), humic acid solution (Wang et al. 2011), water after lime softening (Vahedi & Gorczyca 2011), lake water (Chakraborti et al. 2000; Kim et al. 2001), textile wastewater (Yu et al. 2009; Yu et al. 2012), sewer system wastewater (Zheng et al. 2011). Some attempts were made to characterise alum flocs by image analysis in the drinking water treatment process (Juntunen et al. 2014).
Yu et al. (2005) presented the results of online monitoring of industrial wastewater true colour using digital image analysis and ANN.
An extensive literature search showed an inadequacy of studies concerning the floc-dosage or floc-effluent relations in the coagulation process regarding municipal wastewater or model wastewater.
The fractal geometry of nature and the term “fractal dimension” were defined by Mandelbrot (1967; 1982). Li and Ganczarczyk (1989) described the spatial structure of particles appearing in water and wastewater treatment processes by fractal theory, i.e. fractal dimension. Currently, it is one of the most common ways of flocs characterisation in water and wastewater field. The review of Bushell et al. (2002) summarises the techniques available for measuring fractal properties of flocs and aggregates.
For a 2D projected particle image, the fractal dimension, Dpf defines how the projected area of the particles rises with the perimeter (He et al. 2012):
ܣ ן ܲଶ/ , (1) where A – projected area; P – perimeter of the particles. For the 2D projection of an image, the value of fractal dimension varies from Dpf =1 for the circle shape floc to Dpf =2 for a chain of particles (a line).
The size of an irregularly shaped particle can be determined as the equivalent diameter, dp (He et al. 2012):
݀= (4ܣ/ߨ)ଵ/ଶ . (2) Even though fractal geometry and particles characterisation by image analysis have gained great popularity among researchers, there were not many identifieable attempts towards applying the method for online coagulant dosage control. One of the few examples include research on flocculation control based on the fractal dimension of flocs in the pilot scale of drinking water coagulation (Chang et al. 2005). However, the technique was not yet applied in full-scale.
The weakness of the particles characterisation method, particularly by image analysis, is that some challenges might arise during the image processing. For instance, the need to remove out-of-focus objects has been documented (Keyvani & Strom 2013). Since the particles characterisation by image analysis is based on objects count algorithm, it is important to ensure that the 3D objects (particles) do not overlap in the 2D images and that the number of particles is not under-estimated. Otherwise, the objects’ characteristics could in such cases be over-estimated. Besides, the number of estimated particles on the image depends on the threshold setting and is often a matter of judgment (Bache & Gregory 2007). The problems of image resolution limitations and hardware limitations are gradually being reduced with rapid development in the industry. However, the “ready to use” online solutions with feasible and robust integrated digital image analysis systems are not yet available in the market.
The Grey level co-occurrence matrix (GLCM) method has found broad application in the food industry, for instance, in the determination of meat quality (Shiranita et al. 1998), baking experiments on wheat baguettes (Kvaal et al. 1998) and surface texture characterisation of an Italian pasta (Fongaro & Kvaal 2013). Other applications of texture image analysis for evaluation of food qualities are summarised in a review by Zheng (2006). The studies on colour
texture classifications can also be found in the literature (Palm 2004; Khan et al. 2015; Gui et al. 2013).
Angle measure technique (AMT) is a method of calculating the complexity of an image.
It was first introduced by Andrle (1994) for characterising the complexity of geomorphic lines.
The purpose was to detect changes in coastline complexity as a function of scale. Esbensen et al. (1996) introduced AMT in chemometrics for general applications and use. Later the technique has been successfully used with image analysis applications (Kvaal et al. 1998;
Huang & Esbensen 2000; Huang & Esbensen 2001; Kucheryavski 2007; Mortensen &
Esbensen 2005; Dahl & Esbensen 2007; Fongaro & Kvaal 2013; Fongaro et al. 2016).
Numerous computer vision applications use texture analysis to perform automated image recognition, classification and segmentation (Haralick et al. 1973; Zheng et al. 2006;
Bharati et al. 2004). Texture is a loosely defined term without an accepted or universally quantitative meaning. A general description is that texture is a combination of repetitive patterns of pixel variations organised in some structural way (Russ 2011). Texture can be defined as a measure of the image’s surface roughness defined by parameters of brightness, colour, shape and size variations within some region and its repetitiveness. The texture properties of the materials/objects in the image may be found or correlated in some way. Such properties might be geometric structure, orientation, coarseness, smoothness, roughness and periodicity. Texture is a pattern that can be completely distinct or completely random.
Furthermore, texture could be isotropic (without any preferred orientation) or anisotropic (has definite pattern structure) (Levine 1985; Gonzalez & Woods 2010).
Considering the variability of texture properties and their combinations, an infinite number of textures exist and it is difficult to identify and describe all these structures.
Nevertheless, different databases have been collected to perform computer vision analyses.
One of the most known, classical, databases was collected by Brodatz (1966) and consists of 112 photographs captured under controlled lighting conditions. The data set is commonly used for testing new texture analysis methods, in computer vision and signal processing. Normalized and coloured Brodatz texture data sets are also used (Abdelmounaime & Dong-Chen 2013).
So far, the water and wastewater industry has not employed texture image analysis methods. However, texture analysis has been broadly used for segmentation of the urban scenes (Conners et al. 1984), in situ powder characterisation (Huang & Esbensen 2000; Huang &
Esbensen 2001), grain size characterisation (Dahl & Esbensen 2007), classification of the tree barks (Palm 2004) and identification of uranium ore (Fongaro et al. 2016).